The sample size of 50 was determined based on data availability, recruitment feasibility within the study period, and the exploratory nature of this preliminary investigation. A post-hoc power analysis was performed with G*Power (version 3.1) for multiple regression (α = 0.05, effect size f² = 0.25), which indicated a power of 0.78 to detect medium effects—a value considered acceptable for exploratory longitudinal studies. While a formal power analysis suggested a larger optimal sample size for path analysis, our sample size is comparable to that of similar longitudinal pilot studies and yields preliminary insights into the dynamic relationships examined in this study (13, 14).

The inclusion criteria were as follows: (1) first diagnosis of breast cancer, (2) aged ≥18 years, (3) underwent modified radical mastectomy for breast cancer, (4) normal communication and comprehension abilities, and (5) ability to use electronic devices (e.g., computers, mobile phones) for Internet access (patients or their family members). The exclusion criteria were as follows: (1) combined with other cancers, (2) cancer recurrence or metastasis, and (3) severe cardiopulmonary, hepatic, or renal dysfunction or end-stage disease. This study was conducted in accordance with the Declaration of Helsinki and was approved by the Ethics Committee of The First Affiliated Hospital of Nanjing Medical University (approval no.: 2023-SR-770). Written informed consent was obtained from all participants prior to data collection.

This prospective longitudinal study was conducted in the Breast Surgery Wards of a university-affiliated grade III class A hospital. The participants were recruited postoperatively after their condition had stabilized (usually on postoperative days 2 to 3). Trained research staff explained the study purpose, procedures, potential risks, and benefits to the participants and emphasized to them that participation was voluntary. Written informed consent was obtained from all patients who agreed to participate.

Data were collected electronically via Questionnaire Star platform at three time points: pre-discharge (T1, 3 to 4 days post-surgery), 1 month post-surgery (T2), and 3 months post-surgery (T3). T1 assessment was completed in the hospital using a tablet device, while T2 and T3 assessments were conducted via encrypted electronic links sent to participants’ mobile phones following a reminder call with brief guidance. Brief verbal guidance was provided before each follow-up assessment to ensure data quality.

To minimize potential selection bias, data collection was conducted by trained research staff who were blinded to the study hypotheses and not involved in the clinical care of the participants. All participants were assigned a unique identification code to ensure anonymity during data collection and analysis.

At T1, general demographic characteristics, QoL, functional exercise compliance, FCR level, and PSQI score were collected. The same indicators were reassessed at T2 and T3. A total of 54, 52, and 50 participants were included in the T1, T2, and T3 assessments, respectively. Four participants were lost to follow-up (two declined to answer phone calls and two withdrew from the study), resulting in an effective follow-up rate of 92.6%.

Statistical analysis was performed using SPSS 26.0 and Amos software. Categorical variables were described as frequencies and percentages. Normally distributed continuous variables were expressed as mean ± standard deviation (SD), while non-normally distributed continuous variables were summarized as median (25th, 75th percentiles). Scores for QoL, functional exercise compliance, FCR, and PSQI were normally distributed at all three time points. The assumption of sphericity was violated for functional exercise compliance scores, and the Greenhouse–Geisser correction was applied for subsequent analyses. Repeated-measures analysis of variance (ANOVA) was used to analyze within-group changes across the three time points.

Generalized estimating equations (GEE) were used to identify factors influencing quality of life, accounting for within-subject correlations in longitudinal data. An unstructured correlation matrix was used for flexibility. The model included time, repeated measures (exercise adherence, sleep quality, and FCR level), and time-invariant covariates (demographic characteristics, cancer stage, and adjuvant therapy). Multicollinearity was assessed using variance inflation factors (VIFs), with a VIF <10 indicating no significant multicollinearity. Model fit was evaluated using the quasi-likelihood under the independence model criterion (QIC). Missing data from the four lost-to-follow-up cases were handled under the missing completely at random (MCAR) assumption within GEE, which provides unbiased estimates under MCAR.

Structural equation modeling (path analysis) was chosen as it allows for the simultaneous testing of multiple direct and indirect effects, which aligns with our study’s aim to examine the complex mediation relationships among FCR, exercise adherence, sleep, and QoL. Path analysis was conducted in Amos using FIML estimation, robust under MAR assumptions. Model fit was assessed using the chi-square/degrees of freedom ratio (χ²/df), goodness-of-fit index (GFI), comparative fit index (CFI), normative fit index (NFI), and root mean square error of approximation (RMSEA). The robustness of the path coefficients was evaluated using the bias-corrected bootstrap method with 1,000 resamples, which is recommended for assessing stability in samples of modest size. A p-value <0.05 indicated statistical significance.

A total of 50 patients were included in this study and completed all three longitudinal assessments. The demographic details of the participants are outlined in Table 1.

Data on clinical characteristics including cancer stage (II: n = 34, III: n = 16) and treatment modalities (chemotherapy: n = 31, radiotherapy: n = 29) were also collected but are not presented here due to limited variability in this cohort.

The results showed that patients had the lowest QoL and highest FCR scores at T1, indicating the poorest QoL and the most severe psychological distress related to FCR at this time point. The QoL and FCR scores differed significantly across the three time points (all P < 0.05). Patients had the highest functional exercise compliance and the lowest PSQI scores (best sleep quality) at T2 (Table 2).

The quality of life scores of patients was considered the dependent variable, while general demographic information, as well as the functional exercise adherence of patients (T1–T3), Pittsburgh Sleep Quality Index, and fear of cancer recurrence score were considered the independent variables; all variables were incorporated into a generalized estimating equations model. Variable assignment is shown in Table 3. The results of the study showed that factors such as age, BMI, occupational status, daily working hours, daily pace, exercise intensity, exercise time, functional exercise adherence, fear of cancer recurrence, and the Pittsburgh Sleep Quality Index influenced the quality of life for patients. It is noteworthy that all sociodemographic and clinical variables listed in Table 3 were included in the generalized estimating equation model. However, to enhance the clarity and focus of the results, Table 4 presents only the variables that demonstrated a statistically significant association with quality of life (P < 0.05). Multicollinearity diagnostics confirmed that all time-invariant covariates had variance inflation factor (VIF) values well below the threshold of 10 (range: [1.252 to 3.972]), indicating that multicollinearity was not a substantial concern in the model. The results are shown in Table 4.

This study adopted structural equation modeling to explore the determinants of quality of life at T3, with functional exercise compliance, fear of cancer recurrence, and the Pittsburgh Sleep Quality Index as exogenous variables. The initial saturated model (χ² = 0, df = 0) was not suitable for model fit evaluation (Figure 1). After removing the non-significant path from Pittsburgh Sleep Quality Index to functional exercise adherence (P = 0.357), the revised model showed excellent fit: χ²/df = 0.842 (P = 0.359), RMSEA = 0.000, GFI = 0.992, NFI = 0.980, CFI = 1.000. Bootstrap analysis (1,000 samples) confirmed the robustness of all significant paths, with the bias-corrected 95% confidence intervals excluding zero (Figure 2).

Path analysis using standardized coefficients revealed that fear of cancer recurrence had a significant direct negative effect on quality of life (β = −0.920, P < 0.001) and a significant indirect effect mediated by reduced functional exercise adherence (β = −0.190, P = 0.011). The total effect of fear of cancer recurrence on quality of life was β = −1.086. The Pittsburgh Sleep Quality Index had a significant direct negative effect on quality of life (β = −1.139, P = 0.003), whereas functional exercise adherence had a marginally significant direct positive effect on quality of life (β = 0.874, P = 0.052) (Table 5).

The model exhibited excellent fit indices; however, the RMSEA of 0.000 may indicate potential overfitting, as will be discussed subsequently.

Quality of life (QoL), a multidimensional indicator encompassing physiological, psychological, and social functioning, is a key outcome for evaluating breast cancer rehabilitation. This study found QoL scores of 87.44 ± 11.32 pre-discharge (3 to 4 days post-surgery), 104.16 ± 8.94 at 1 month, and 100.5 ± 11.02 at 3 months. Although a slight decline occurred relative to the 1-month peak, QoL at 3 months remained higher than the pre-discharge levels, indicating an overall trend of gradual improvement. As supported by Zhang et al. (21), significant functional recovery and adaptation to surgical trauma were observed after 1 month. With the passage of time, patients resumed daily life with reduced psychological stress, which contributed to the sustained improvement of QoL. Although a minor decrease occurred at 3 months (possibly related to work and lifestyle transitions), QoL remained substantially improved compared to the early postoperative phase. These findings suggest that early postoperative care should prioritize physical rehabilitation, with the aim of facilitating the restoration of bodily functions in a timely manner, while later stages require psychosocial support to promote social reintegration and sustained QoL improvement.

This study shows breast cancer patients with postoperative functional exercise adherence, sleep quality, and fear of cancer recurrence in a dynamic change (22). Exercise adherence was low in the early postoperative period, which may be attributed to surgical pain, fatigue, and treatment-related adverse effects. As rehabilitation progressed, exercise adherence improved significantly; this finding is inconsistent with that of Chu et al. (23), which may be due to the implementation of individualized rehabilitation programs and regular follow-up in their study. Sleep quality followed a fluctuating pattern: lowest at T1, improved at T2, and mildly declined again at T3. The initial decrease may relate to chemotherapy’s adverse effects, while T2 improvement may reflect psychological adaptation. The T3 rebound may be associated with fear of recurrence and long-term stress, consistent with the results of Lu et al. (24). Fear of cancer recurrence after surgery was the first to fall fast, within 3 months after the slow decline of linear trajectory. This is consistent with the research findings of Dunn et al. (25). Early postoperative recurrence in patients with fear is relatively strong, which may be related to the uncertainty of the surgical trauma and disease prognosis. As time goes by, the patients gradually accept their disease status and adapt to daily life, leading to a significant reduction in fear of cancer recurrence. Postoperative recovery is characterized by early physiological limitations, mid-term psychological and sleep improvements, and long-term stabilization. Clinical interventions should be stage-specific: early emphasis on pain management and exercise guidance, mid-term integration of psychological support and sleep interventions, long-term focus on maintaining healthy behaviors, and managing fear of recurrence to optimize a holistic recovery.

The results indicated that age, BMI, occupational status, daily working hours, daily gait speed, exercise intensity, and exercise time were significant factors influencing quality of life (QoL) in breast cancer patients. Specifically, elderly patients had lower QoL, which is consistent with the findings of Li et al. (26). Possibly due to decreased physical function and higher comorbidity burden, highlighting the need for enhanced health management and psychological support in this group. Higher BMI was also associated with reduced QoL, aligning with Carolyn et al. (27). Obesity may increase body burden and treatment side effects, underscoring the importance of weight management interventions (28). Occupational status and longer daily working hours significantly affected QoL, likely reflecting the dual burden of work and health management. The work-related stress and overall life burden of younger patients are higher, and related research results (29) indicate that low working pressure helps reduce psychological burden and promote rehabilitation. In addition, patients with a faster daily pace, higher intensity, and greater levels of activity tend to maintain a higher quality of life for a longer time, according to research results such as that of Burse et al. (30). Exercise can improve physical function and reduce treatment-related adverse effects. Thus, it is recommended to integrate exercise interventions into the routine postoperative rehabilitation program of breast cancer patients.

Path analysis revealed complex longitudinal associations among exercise, sleep, psychological fear (FCR), and QoL in breast cancer patients. While the temporal sequence provides evidence for potential mechanisms, the observational nature of our design necessitates a cautious interpretation of these relationships as associative rather than strictly causal.

Sleep quality, psychological state, and exercise engagement interact through multiple, interrelated pathways. Poor sleep contributes to daytime fatigue and reduced energy, which directly limits the patients’ exercise participation (31, 32). Similarly, fear of recurrence can lead to kinesiophobia, where patients avoid physical activity due to unwarranted safety concerns, further reducing exercise adherence (33). Both sleep disturbances and psychological distress also diminish self-efficacy, impairing their motivation to engage in regular exercise (34, 35).

These behavioral interactions are likely supported by underlying biological pathways. On a neurobiological level, regular exercise is known to promote the release of endorphins and reduce cortisol levels (36), which can alleviate psychological stress and improve mood. Improved mood and reduced stress, in turn, can lower hyperarousal and facilitate better sleep quality (37). Conversely, fear and poor sleep can dysregulate the hypothalamic–pituitary–adrenal (HPA) axis, increasing inflammation and fatigue (38). This physiological state creates significant biological barriers to exercise participation, forming a vicious cycle that ultimately impairs the overall quality of life.

Consequently, functional exercise compliance was positively associated with QoL, likely through mechanisms such as enhancing physical function and immunity, a finding consistent with prior research (30). The Pittsburgh Sleep Quality Index was negatively associated with QoL; sleep disorders may exacerbate fatigue, pain, and mood disturbances, potentially disrupting neuroendocrine and immune function (31, 32). Fear of recurrence adversely affected QoL through dual pathways: it directly impaired coping ability (33) and indirectly reduced QoL by decreasing exercise adherence and disrupting sleep patterns (34, 39).

Several important limitations of this study must be considered when interpreting these findings. First, the relatively small sample size (n = 50) from a single center limits the statistical power and generalizability of our results. Although we employed bootstrap resampling to enhance the robustness of our path estimates, the modest sample size may still affect the stability and precision of the structural equation model. The use of convenience sampling may also introduce selection bias. Second, although we employed a longitudinal design, the presence of unmeasured confounding variables cannot be ruled out. Furthermore, the excellent model fit (RMSEA = 0.000) should be interpreted with caution, as it may reflect overfitting to our specific sample rather than true population dynamics. Finally, the use of self-reported measures, though validated, may be subject to reporting bias. These limitations highlight the need for cautious interpretation and further validation in larger, more diverse populations.

In summary, the findings of this exploratory study are consistent with a sequential pathway wherein improved sleep may enhance psychological well-being, which could facilitate exercise adherence, and ultimately contribute to improved QoL (35). Specifically, better sleep alleviates symptoms of anxiety and depression while also strengthening psychological resilience (35). Better psychological health can enhance self-efficacy, encouraging participation in functional exercise (35). In turn, regular exercise not only improves physical function but also reduces psychological stress through endorphin release and other mechanisms, establishing a virtuous cycle. Clinically, exercise therapy might play a central role by improving physical function, reducing stress, and enhancing sleep, thereby disrupting the pathological cycle (40). Combined psychological and exercise interventions have been shown to improve sleep and reduce fear (41). Based on our preliminary findings, we suggest that postoperative management could consider incorporating exercise, sleep regulation, and psychological support into multidimensional rehabilitation strategies as potential avenues to holistically improve QoL.